Subproject 3

Patrycja Jabłońska

Temporal relations across sentences and languages:
A case study of temporal converbs

The main research question

Whereas SP-1 and SP-2 investigate which categories are responsible for temporal relations within one clause (→ intrasentential domain of investigation), the perspective taken in SP-3 is different, namely, the focus will be on temporal relations between main and subordinate clauses (→ intersentential domain of investigation). The sort of subordinated clauses we will look at are converbs (also referred to as conjunctive participles, gerunds, depending on the linguistic tradition). These are temporally anchored participial clauses functioning in a sentence as optional (i.e., adjunct-like) adverbial modifiers. From the point of view of the temporal relations between the matrix and the embedded event three types of converbs can be distinguished: (i) simultaneous converb (e.g., Polish -ąc-participles); (ii) anterior converb (e.g., Polish -wszy/łszy-participles); and (iii) posterior converb (e.g., Shipibo-Konibo converb; cf. Velanzuela 2005); cf. (1).

(1) a. simultaneity (EVENT-1 TIME coincides with EVENT-2 TIME)
anteriority (EVENT-1 TIME precedes EVENT-2 TIME)
posteriority (EVENT-1 TIME follows EVENT-2 TIME)

Given that ‘posteriority’ might be conceived of as a type of future reference, and that ‘posterior converbs’ are quite rare typologically, there arises an interesting correlation with the marked status of future tense mentioned in SP-1 and SP-2.

In spite of the fact, however, that the relations in (1) are temporal in nature, it is far from clear that it is the category Tense that is responsible for establishing them. This is due to the controversial categorial (or ‘word class’) status of converbs (see Topic 2 below).

Working hypothesis

In accordance with SP-1 and SP-2 we hypothesize that the expression of temporal relations between the main and the converb clause might be a result of the interactions of multiple categories, possibly even categories other than Tense. That this initial hypothesis might be on the right track seems to be supported by the existence of English converb pairs such as doing vs. having done, which differ mainly in their aspectual specification, nevertheless triggering distinct temporal readings. Similarly, Polish displays aspectual restrictions w.r.t. converb formation (simultaneous converbs with imperfective verbs only, anterior converbs with perfective verbs only). This makes us address the following more specific topics:

Topic 1: The morpho-syntactic ‘make-up’ of converbs across languages and its role in establishing temporal relations

Traditionally, converbs have been conceived of as non-finite verb forms. It is a far from trivial issue to decide what ‘non-finite’ really means on a cross-linguistic scale, but let us assume that NON-FINITE = TENSELESS. That hypothesis results in a prediction that it should be impossible to find converbs with a morphological manifestation of the category of Tense. Our initial survey indicates that this might in fact be true. If this is correct, other questions arise, namely:

More specific research questions
Q1: How is it possible that converbs apparently express temporal relations? Which other categories conspire to give temporal meanings? In other words, what triggers temporality in the absence of tense morphology?

Q2: What is the exact semantic contribution of the individual morphological exponents present in converbs? For example, what is the actual semantic import of the participial
suffix -l in Polish anterior converbs (see SP-1 and SP-2) or the so-called thematic suffix on Polish verbs in the context of the categorial status of converbs (see Topic 2 below)? How do certain types of morphomic syncretism (cf. Aronoff 1994) (or ‘parasitic’ syncretism (cf. Mathews 1972)) arise, e.g., the fact that the two Polish converbs are built on two different stems respectively?

Q3: How much syntactic structure is hidden under the visible morphology of converbs? Is it possible that converbs in different languages vary as to how much “phrasal” structure they contain? In other words, do we expect to find more “clause-like” or more “word-/phrase-like” converbs? (e.g., in the sense of different cut-off points on the scale of desententialization; see Lehmann 1988)?

Topic 2: The categorial status of converbs cross-linguistically

It is also a matter of debate what categorial status converbs have. In view of the fact that the cut-off point between features present/absent in a converb can lie at various places in a continuum (see Koptjevskaja-Tamm 1994 for some relevant discussion), treating converbs as a separate word class on its own (as in Meščaninov 1945) is questionable. The next logical possibility would be to classify converbs as belonging to the word class category VERB. However, this cannot be entirely correct given the existence of NOUN-like converbs (e.g., in Turkish, Ge’ez, arguably Evenki, in which converbs bear inflection characteristic of unambiguous nouns) or ADJECTIVE-like converbs (e.g., in Panoan or Australian languages, in which converbs display participant agreement (with the Ergative- or Absolutive-marked argument of the matrix clause (see Velanzuela 2005)). Similarly, in Ancient Greek the so-called participium conjunctum displayed Gender (and Number) agreement as did typical adjectives.

More specific research question
Q4: Given the cross-linguistic variation and disagreement among linguists as to the categorial status of converbs, the question arises whether we should assume discrete categories or whether it is more correct to understand categories in a “scalar” sense (e.g., something can be more/less nominal or verbal). If the former, then what criterion is critical in defining certain sets of features as ‘verbal’, ‘nominal’ ore ‘adjectival’? One hypothesis is inflectional marking (e.g., person agreement for verbs and gender agreement for adjectives; cf. also Baker 2008). The other one is conceptual semantics of a given root (as in the semantic map approach to participant-oriented and event-oriented adverbs in Himmelmann and Schultze-Berndt 2005). In this project we will try to tease apart the ramifications of the two theories using novel experimental evidence (see below).


As far as Topic 1 is concerned, in light of the fact that the semantic contribution of particular morphological exponents is extremely difficult to evaluate due to massive polysemy, a cross-linguistic comparison between features present/absent in a converb must be preceded by a thorough language-internal investigation. To give a specific example, it could potentially be argued that apart from aspectual distinction, past tense morpheme is also present in the Polish anterior converb (przy-szed-ł-szy ‘prefix-go-PST-CONV_ANT’). Yet, as has already been mentioned in SP-2, absolute past tense is just one of the many contexts where the exponent is used, e.g., it is also present in optative mood (Oby przyszedł! ‘May he come!’), which can hardly be construed as containing (whether absolute or relative) past tense. Therefore, a thorough investigation of a particular exponent’s role in the system has to be carried out (see also Q2).

In order to answer the question posed in “Topic 2” we will conduct experiments to make our later theorizing psychologically real. This will help us to verify the hypotheses advanced by the project.

Innovative character of SP-3

The most innovative feature of this subproject is that its theoretical part will be supplemented with neurolinguistic evidence. There is converging body of neurolinguistic evidence suggesting that the grammatical category distinction VERB vs. NOUN is essential for production, as well as Event Related Potentials (ERP). As shown by Cappelletti et al. (2008), repetitive TMS applied to the anterior portion of the left midfrontal gyrus interfered only with the ability to produce verbs, but not nouns. Similarly, Kellenbach et al. (2002) in a series of experiments managed to tease apart attribute features (e.g., manipulable vs. abstract) from grammatical class effects. The latter included: (i) posteriorly maximal enhancement of the P2 component evoked by verbs relative to nouns and different neural populations contributed to ERPs elicited by nouns and verbs in that time interval; (ii) greater positivity elicited by verbs over the midlatency range 350-450 msec (N-400).

Most of the existing literature, however, involved comparison between finite verbs and nouns. We would like to run a series of ERP experiments to verify whether the two types of Polish converbs would elicit the relevant components. The comparison would be between converbs and full-fledged clauses with finite verbs on the one hand; and converbs and nominalizations (the so-called nomina deverbalia) on the other.